docs(glossary): complete agent/context/eval/templates/tokenwise terms

[userName20260323]

Change description

Complete the domain glossary (CONTEXT.md) for the five modules under my review — Agent Core (agent/), Context Engine (context_engine/), EvalEngine (eval_engine/), Workspace Templates (templates/), TokenWise (token_wise/) — and fix one stale entry. Branched off the review baseline docs/glossary_baseline_and_init_docstring per the team-review workflow; additive only, no other owner's terms touched.

context_engine — rewrite the stale "Context Engine" entry: the historical legacy / curator / default engine split was collapsed into one ContextAssembler (the engine: setting is now only a backward-compat alias); add SegmentBuilder, Segment, Prompt Segments (# Memory / # Skills / # Active Skills), ContextPlan; note the Curator is now Segment 6.

agent — add Tool, Tool Registry, Checkpoint, Empty-Response Recovery, Synthesis, Personalizer, Context Builder. huangjie's Agent Loop / Turn Runner / Agent Hook / Subagent entries are left untouched.

eval_engine — add EvalJudge and JudgeVerdict (the actual task-completion judge and its three-state verdict).

templates — add a "Workspace & Onboarding" section: Workspace, Workspace Template, Onboarding, Bootstrap Files.

token_wise — add StrategyRegistry, UsageTracker, CacheOptimizer, UsageSnapshot, and the before_llm_call / after_llm_call hook seam (TokenWise / TokenStrategy verified accurate, refined only). Planned-but-unshipped strategies in PLAN.md are intentionally omitted.

Out-of-scope findings surfaced during the same pass (two accuracy nits in huangjie's Agent Core entries, plus a package docstring gap) were sent to the team separately and are intentionally not changed here.

Type of change

• Bug fix
• New feature
• Document
• Others

Related issues (if there is)

Refines the domain-glossary review baseline (PR #4 / glossary finalize). No standalone issue.

Checklists

Docs-only change (glossary text in CONTEXT.md); no application code, so the Development and Security checklists below are not applicable.

Development

• Lint rules pass locally
• Application changes have been tested thoroughly
• Automated tests covering modified code pass

Security

• Security impact of change has been considered
• Code follows security best practices and guidelines

Code review

• Pull request has a descriptive title and context useful to a reviewer. Screenshots or screencasts are attached as necessary